Base Stoppers

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I notice that base stoppers may help to stop oscilation.

What is the mechanism with this 10R-100R in series with base that makes it stops oscilation? What is wrong with transistor's plain base?

When searching here, some members explaining about "high impedance" meeting "low impedance" terms, but I don't get it. How to understand this "impedance" terms more easily?

Some designs "Replace" this base resistor with a B-E resistor about 470ohm (without series base resistors here).

Is this 470R B-E resistor stops oscilation by shooting at the same cause (that series base resistor targets)?
 
RC filter can be used to limit higher frequency.

MOSFETs have an input capacitance.
Gate resistor provides the R and so we have RC filter.
Gate resistor or Base resistor should be put as close as possible
to Gate or Base.

This way we can stop high freq oscillation, if we need.

The higher value of R, the lower this filters cut off freq will be.
 
Help me! Please!

Hi everyone,

I have a similar question. I have constructed an amplifier (schematic here), and was surprised to discover that the output stage suffers from parasitic oscillation (surprised since it is an EF, and I didn't think they suffered from parasitic oscillation, and there are base-stopping resistors). The oscillation is at approximately 3.57 MHz with an amplitude of just over 5 Volts peak to peak.

Please note the following:

The power supply is ±45 V regulated (coming from a TTi CPX400A bench P.S.U.)

The output stage suffers from parasitic oscillation even with only one pair of output transistors (2SA1302/2SC3281) connected.

With no output transistors connected, no load connected (apart from oscilloscope) and the junction of R17 and R28 (i.e., the output junction) connected to the junction of Q14 emitter and R16 (i.e. feedback applied around input stage, VAS and drivers), the amplifier works as expected with no unwanted oscillations. This suggests a problem with the output stage

With all output transistors connected, but with the feedback point moved to the junction of D4 and D5 (i.e. running the amplifier with the output stage open-loop), the oscillation is still present.

MBR150 diodes are schottky types with a Vf of 0.3 V at the VAS operating current. In conjunction with the 1N4148 diodes in the bias chain, they result in a Vq across the output resistors of around 80 mV, resulting in an Iq of about 364 mA per output device. There is a 47 uF cap across the diodes which I forgot to put in the schematic.

The amp. is DC coupled as I'm not intending to use this amplifier for audio. Rather, it is destined to become part of my test equipment and needs to be able to output arbitrary waveforms that may or may not have DC components.

If any one has any comments on this, please let me know. I hope that any insights you have will also assist lumanauw.
 
In one thread "Miller or not miller compensation", some member said an interesting thing.

At a certain frequency, the transistor base acts like a inductor, and this base stopper gives real impedance to the base.

What is the meaning of this? How to understand it easily?

The predriver is usually emitor follower (common collector). Also from the same thread, this predriver needs base stoppers because the VAS output (collector) is low impedance.

I cannot imagine what is low and high impedance. VAS collector gives low impedance? Is there an easy way to understand this?
 
Hi, HarryDymond,

I'm not sure, but with oscilation you must try and error

-put 1k in base of Q3
-change the double pole (c6,c7,r44) to ordinary single pole compensation. Search for value
-before Q24, you can try putting a buffer (emitor follower), because you use current mirror in differential
-try feedforward cap, ie : pf cap from VAS output to inverting input of differential
-put base stoppers infront of Q6-Q14, values up to 470R or put 470R between those B-E (This is what I asked :D)
-Split R16 to 2x100ohm, with junction connected to output, if you are biasing the output stage in classAB.
 
You could try to put resistor in supply rail to VAS/input,
maybe 10-47 Ohm between R2 and collector of Q6.
And add a capacitor 100uF or so from the R2 side to ground.
Same for the negative rail.

In this way we protect input/VAS to be effected by supply variations
caused by the output stage.
Stop unwanted 'feedback' to reach input stage.

I dont know if this is a problem,
but you use very high value resistors in feedback, 91k / 10 k.
Same to set the input impedance. (= 100k )
This is not often seen.
 
Hi Harry,

Remove R25 & C3, this will run the amp open loop. Check to see if your output is stable. If not, then you are dealing with a parasitic. If it is stable, then you are dealing with a loop compensation problem... different beast.

You can then maesure the gain and phase response of the amp. Can you measure the phase response? Need a primer??? You will need a sine wave generator that will go to 10 mHz and a scope...
 
Hi everyone, thanks for all your replies. I won't be back to university until Jan 3, so I'll have to wait until then to try some of these things out...

Unfortunately, I was a bit 'gung ho' and built this amp straight to PCB, so that may limit what I can try without writing the whole thing off and having to start again :(


lumanauw said:
-change the double pole (c6,c7,r44) to ordinary single pole compensation. Search for value

I'm 99% sure that this is parasitic oscillation in the output stage. If I put a sine wave at the input, I get an amplified sine wave out, but with this 3.5 MHz oscillation 'piggy-backed' on top. If there was a problem with the compensation, the amp would just oscillate and not amplify anything, no?


lumanauw said:
-before Q24, you can try putting a buffer (emitor follower), because you use current mirror in differential

I'm using a BC550C here (min hFE = 420), which in combination with the 33R degeneration resistor provides a high input impedance.


lumanauw said:
-try feedforward cap, ie : pf cap from VAS output to inverting input of differential
-put base stoppers infront of Q6-Q14, values up to 470R or put 470R between those B-E (This is what I asked :D)
-Split R16 to 2x100ohm, with junction connected to output, if you are biasing the output stage in classAB.

As I said, the amp is already on PCB (and the input stage, VAS and drivers are squeazed in to a small PCB area) so I'll have to resort to these ideas if I am forced to give up on my current PCB.


mzzj said:

Ferrite beads are sometimes also used on gate for same purpose. Maybe more common in SMPS than amps tho. Choosing right one is more tricky and availlability is not nearly as good as resistors..

In February this year, when I went to the Bristol HiFi show, NAD had some amps on display with the lids off. Their output transistors had ferrite beads around one leg. Stupidly, I didn't think to note down the transistor number or which leg it was... doh! Does this not also increase the inductance of the leg around which the bead is placed? Can you suggest any books with good sections/information about ferrite beads?


AndrewT said:
Hi,
what if the output stage is not causing the problem?

Could the regulated lab type PSU be the source of the oscillation?

Andrew, you make a good point. I was wondering this myself. The PSU is a switching/linear hybrid, and I was using 0.5 meter long leads to connect it to the amp PCB. The PCB has its own standard bridge & large capacitor power supply, but I like to test stuff with a current- limited power supply first.

I have built a separate amplifier, for audio use (schematic alterations - not shown is RC filtering on the negative rail to the VAS and current mirror, and the two-pole compensation actually feeds-back to the junction of Q8 collector and Q12 emitter), which I tested with the same PSU without problems.


lineup said:
You could try to put resistor in supply rail to VAS/input,
maybe 10-47 Ohm between R2 and collector of Q6.
And add a capacitor 100uF or so from the R2 side to ground.
Same for the negative rail.

In this way we protect input/VAS to be effected by supply variations
caused by the output stage.
Stop unwanted 'feedback' to reach input stage.

The power supply rejection on the +ve rail should already be high (but I suppose perhaps not at higher frequencies?). I could try adding RC filtering to the negative rail.


lineup said:
I dont know if this is a problem,
but you use very high value resistors in feedback, 91k / 10 k.
Same to set the input impedance. (= 100k )
This is not often seen.

Good point! I'm sure I had a reason for that. If only I could remember what it was! I'll try lowering these values.


poobah said:
Remove R25 & C3, this will run the amp open loop. Check to see if your output is stable. If not, then you are dealing with a parasitic. If it is stable, then you are dealing with a loop compensation problem... different beast.

I can't tell if you're joking or not? The gain of this amp open loop is huge, surely it would just saturate if I tried to run it completely open loop?

I would have thought that the things I tried as outlined in my first post, especially running the amp with the loop closed around the input stage/vas, but leaving the output stage outside the loop, suggest that this is a parasitic oscillation in the output stage.

Thanks once again to everyone who replied.
 
Harry,

most of these posts here are about stabbing around and playing capacitor roulette.

Now... capacitor roulette is fine, I play it too, but you have to start from the standpoint of a stable amp and work up to the point of maximum loop gain incrementally. There are two ways to do this.

1. Back way off on the loop gain and play roulette. (double, triple, or more on R25.

2. Measure the gain and phase over frequency of the forward function input to output R25 & C3 OUT...

Use 30 Hz, 100 Hz, 300 Hz, 1000Hz and so on till you get to 10 mHz.

You are right, it is not likely that Emitter follower are causing the problem.

Choose roulette or measure, and I'll give you the steps.

:D :D
 
Are your 2SC3281 and 2SA1302 all fakes? all genuine devices? maybe a mix of them?

Note that these devices were discontinued a long time ago and all what is sold now with these part numbers are just fakes with quite random behaviour.

Also, these output transistors are quite fast, so if they were genuine they wouldn't probably oscillate at 3.5Mhz but above 10Mhz (I've experienced this with Sankens, even at 30Mhz). Note that 3.5Mhz is still in the frequency range where most circuits employ global negative feedback, so it may be actually a global loop instability issue.
 
poobah said:
2. Measure the gain and phase over frequency of the forward function input to output R25 & C3 OUT...

Use 30 Hz, 100 Hz, 300 Hz, 1000Hz and so on till you get to 10 mHz.

You are right, it is not likely that Emitter follower are causing the problem.

Choose roulette or measure, and I'll give you the steps.

Hi Poobah,

This amp will not work without a closed feedback path. I know how to acquire the data for constructing a bode plot of a circuit, but I think the amp just won't work open loop so I won't be able to make any measurements. What was the procedure you were thinking of?

I would suggest that if I wanted to know the theoretical open-loop response of this amp, I could increase the closed-loop AC gain to - let's say - 1000, then observe the difference voltage at the inverting and non-inverting inputs. I could then calculate open loop response from the closed loop response of the error voltage.

Also, you seem pretty convinced this is a loop stability issue, despite the fact that the oscillation is present even when the output devices are not in the loop, and that it goes away with the output devices disconnected.


Eva said:
Are your 2SC3281 and 2SA1302 all fakes? all genuine devices? maybe a mix of them?

Note that these devices were discontinued a long time ago and all what is sold now with these part numbers are just fakes with quite random behaviour.

Also, these output transistors are quite fast, so if they were genuine they wouldn't probably oscillate at 3.5Mhz but above 10Mhz (I've experienced this with Sankens, even at 30Mhz). Note that 3.5Mhz is still in the frequency range where most circuits employ global negative feedback, so it may be actually a global loop instability issue.

See my thread on the issue of whether the devices are fake or not.

Also, note that I have a different amplifier (schematic linked in my previous post) using 2SA1302/2SC3281 devices from the same source (RS) and this is working without any problems.

Interestingly, at first I had made an error in my bias chain and the bias voltage was higher than I wanted (about 200 mV Vq across the output resistors). With this bias level, the oscillation was also present, but of a much higher frequency.
 
Harry I guess I don't understand why it won't run open loop... schematic as shown, with only R25 & C3 removed. I miss stuff all the time... ask EVA... can you elaborate?

Also, you cannot rule out loop gain just because of the output transistors... I don't know the device specs... but they could contributing phase shift at 3.5 mHz.

I guess I don't under stand about the ouputs being "out of the loop" vs. "disconnected". Wish I could get the "quote" thing to work. If I under stand that right I would be looking at the power supply.

If it is output parasitics it should show up with the junction of D4/D5 grounded.

Divide and conquer!
 
poobah said:
Harry I guess I don't understand why it won't run open loop... schematic as shown, with only R25 & C3 removed.

The input differential amp and VAS together have enormous DC gain. Without a feedback path, the amp would probably saturate. And even if it didn't, I'd probably have difficulty producing a small enough AC input signal to test it with.

poobah said:
I guess I don't under stand about the ouputs being "out of the loop" vs. "disconnected". Wish I could get the "quote" thing to work. If I under stand that right I would be looking at the power supply.

A diagram would certainly help me to explain this. Unfortunately I don't have access to a proper schematic program at home, and I'm going away for Christmas tomorrow. I can post an explanatory drawing in the new year.

poobah said:
If it is output parasitics it should show up with the junction of D4/D5 grounded.

Good point. I should try that.
 
I order to test the amplifier open-loop at high frequencies but without losing DC balance, connect a high value capacitor (1,000uF or more) in paralell with R27.

How does this change affect the oscillation?

Note that this tweak also allows to play with inner compensation without interacting with the outer global loop... You may even find out that the output stage is oscillating due to interaction with the miller capacitor...
 
I got ya Harry,

Drive the input with a 1K/10R divider with the cap like EVA says. You can measure your gain then... or at least you get the idea... kepp the lower resistor of the divider super low to avoid noise pick-up.

Remember, gain is great but worthless without phase info... I know it it's no fun... but once you are set up it'll take 10 minutes to run the through the freq's. Make SURE to recheck your input level for each freq; in other words don't include your generator in the results!

I stress this because to separate loop gain from parasitics you will still need to open the loop. It's either that... or play roulette.

I'm sure EVA will have good ideas (smart) to add.

Good holidays for all of you!

:cheers: :cheers: :cheers: :drink: :drink: :drink: :emoticon: :emoticon: :crackup: :dead:
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.